Detecting County-Level Clusters of Mortality Due to Excessive Alcohol Consumption in the United States: A Space-Time Analysis From 1979 to 2007

Background: Mortality due to excessive alcohol consumption is known to vary substantially by US state. The importance of characterizing the spatial variation in alcohol-related (AR) mortality at the county-level for epidemiological research and public health program planning has been recognized for decades. Yet, there has been limited research on spatial clustering of AR mortality at the county-level over an extended period of time. The purpose is to 1) identify county-level clusters of excessive alcohol mortality; 2) to determine if those clusters are persistent through several decades; and 3) to describe the localization of the clustering in both space and time.

Methods: We used the Centers for Disease Control and Prevention’s (CDC’s) all-cause mortality data to count ICD-coded deaths by US county due to conditions 100% attributable to alcohol consumption.  Attributable fractions came from the CDC’s Alcohol Related Disease Impact classification system.  We performed purely spatial analyses using a Poisson probability model in SATSCAN™, in each of three recent decades, using the scanning window set at 5% of the population at-risk.  Rural-urban classification and personal income were used as ecologic covariates.   

Results: The mean number of AR disease deaths in each county was 209, 235, and 260 people in 1979-1988, 1989-1998, and 1999-2007, respectively.  The mean age-adjusted rate in each county was 25.6, 27.4, and 30.4 per 100,000 population by decade, respectively. In 1979-1988, we identified six potential clusters of counties with age-adjusted rates of AR disease mortality higher than 1.5 SDs from the mean rate of 25.6 per 100,000 population.  In 1989-1998, we identified eight potential high-rate clusters.  In 1999-2007, two new clusters were identified, while there was strong stability in the clusters identified in previous decades.

Conclusions: We identified six to ten clusters of counties with elevated higher than expected risks of AR disease in each decade.  Many of these clusters persisted through time.  For example, a 5-county area in South Dakota persisted through all three decades. We also identified a stable cluster in the Four Corner regions, which are geographic areas of Native Americans living on or near reservations with low levels of health care coverage. Findings highlight the need to identify strategies to reduce persistent AR disease in local areas of the United States. 

Key Words: Alcohol-related mortality, Clustering, Time, Time-intensive